image classi?cation; machine learning; image processing; perceptron separation tree

The paper introduces a real-life industrial problem: a jewelry stones classi?cation. The stones are represented by their camera images. The goal of the contract was to evaluate stones into two (or more) speci?ed classes according to their quality. Given requirements include very high processing speed and success rate of the classi?cation. The goal of this paper is to publish a report of this contract and show a way how this task can be solved. In this paper we aim to usage of machine learning with respect to the image processing. We also design own learning and classi?cation algorithm and answer the question if there is a place for a new machine learning algorithm. As an output of this paper a benchmark of the proposed algorithm with 81 state-of-the-art machine learning methods is presented.